I am one person who has been working on this software for a few years. I got a degree in applied mathematics years ago, but never really used it for anything. I work as a chef and do this in my spare time. I decided to teach myself how to code doing something related to mathematics. I needed a project to keep me motivated, which I suppose it got a bit out of hand. I started by building framework using tensorflow that allowed me to work directly with weight matrices more easily. Tensorflow had no way of using sparse weight matrices, which seemed interesting, so I made a new library from scratch. This software is really only intended for me, but the methods that I use could be implemented in other libraries quite easily. I will try to explain them as clearly as possible so that anyone can experiment with them without needing to read through my code.
One thing that might explain why I do things the way I do is that I do everything on my own PC. I have an old 980 Ti, it has 6GB of RAM and overheats in summer. I could use cloud services, but the constraints force me to look at things that other people might not be looking at.
Email me
sparana@protonmail.com (My name is James)
Discord link
@sparanablog